DiSPEL: Distributed Security Policy Enforcement for Bus-based SoC

Sudipta Paria,Swarup Bhunia

2023-08-05

Abstract:The current zero trust model adopted in System-on-Chip (SoC) design is vulnerable to various malicious entities, and modern SoC designs must incorporate various security policies to protect sensitive assets from unauthorized access. These policies involve complex interactions between multiple IP blocks, which poses challenges for SoC designers and security experts when implementing these policies and for system validators when ensuring compliance. Difficulties arise when upgrading policies, reusing IPs for systems targeting different security requirements, and the subsequent increase in design time and time-to-market. This paper proposes a generic and flexible framework, called DiSPEL, for enforcing security policies defined by the user represented in a formal way for any bus-based SoC design. It employs a distributed deployment strategy while ensuring trusted bus operations despite the presence of untrusted IPs. It relies on incorporating a dedicated, centralized module capable of implementing diverse security policies involving bus-level interactions while generating the necessary logic and appending in the bus-level wrapper for IP-level policies. The proposed architecture is generic and independent of specific security policy types supporting both synthesizable and non-synthesizable solutions. The experimental results demonstrate its effectiveness and correctness in enforcing the security requirements and viability due to low overhead in terms of area, delay, and power consumption tested on open-source standard SoC benchmarks.

Cryptography and Security,Hardware Architecture

What problem does this paper attempt to address?

### Problems the paper attempts to solve

This paper aims to address the security challenges faced in modern System - on - Chip (SoC) designs. Specifically, the currently adopted zero - trust model in SoC designs is threatened by multiple malicious entities. Therefore, various security strategies need to be incorporated into SoC designs to protect sensitive assets from unauthorized access. These security strategies involve complex interactions among multiple IP modules, posing challenges for SoC designers and security experts in implementing these strategies and for system verifiers in ensuring compliance. The main problems include:

1. **Difficulty in policy upgrading**: As security requirements change, upgrading existing security policies becomes complex and time - consuming.
2. **IP reuse problem**: When reusing IP modules for systems with different security requirements, security policies need to be reconfigured, increasing the design time and time - to - market.
3. **Increased design complexity**: Modern SoC designs contain multiple IP modules, and each module may have different security requirements, leading to an overall increase in design complexity.
4. **Trust issues with third - party IP**: When using insecure third - party IP modules, how to ensure the overall security of the system is a major challenge.

To solve these problems, the paper proposes a general and flexible framework, called DiSPEL (Distributed Security Policy Enforcement Logic), for enforcing user - defined security policies in any bus - based SoC design. This framework ensures trusted bus operations through distributed deployment of policies, even in the presence of untrusted IP modules. Specifically, DiSPEL relies on a centralized module that can implement diverse security policies and generate the necessary logic, attached to bus - level wrappers, to support IP - level policies.

### Main contributions

1. **Automated framework**: Proposed an automated framework for enforcing system - level security requirements through formally represented synthesizable security policies, applicable to any bus - based SoC design with low overhead.
2. **Centralized policy module**: Enforce bus - level security policies through a centralized policy module while supporting IP - level policies through bus - level wrappers.
3. **Configurable security specifications**: Provide a convenient option that allows end - users to configure security specifications in a high - level format, which can be extended to different bus - based SoC configurations, applicable to specific threat models and trust assumptions.
4. **Experimental results**: The experimental results demonstrate the effectiveness and capability of the proposed process in efficiently achieving SoC security, and the simulation and formal verification results confirm the correctness of the entire framework.

Through these contributions, DiSPEL aims to improve the security of SoC designs, reduce manual workload, enhance response time, and make threat detection and mitigation measures more proactive and effective.